Agents Python Package Monorepo

Midori AI Agents is a collection of Python packages for building LRM agents systems.

---

Getting Started

Installing All Packages

Install the complete suite of Midori AI agent packages with a single command using your preferred package manager:

With UV (recommended):

# Add to your Python project
uv add "git+https://github.com/Midori-AI-OSS/agents-packages.git#subdirectory=midori-ai-agents-all"

# Or install in a temporary environment
uv pip install "git+https://github.com/Midori-AI-OSS/agents-packages.git#subdirectory=midori-ai-agents-all"

With Pip:

pip install "git+https://github.com/Midori-AI-OSS/agents-packages.git#subdirectory=midori-ai-agents-all"

This installs all Midori AI packages along with embedded documentation that you can access programmatically.

---

Package Documentation

Each package has detailed documentation in its docs.md file:

Core Agent Packages

• midori-ai-agent-base - Common protocol and data models for agent backends
• midori-ai-agent-context-manager - Context management and memory persistence
• midori-ai-agent-huggingface - HuggingFace local inference backend
• midori-ai-agent-langchain - Langchain backend implementation
• midori-ai-agent-openai - OpenAI Agents SDK backend

Intelligence & Processing

• midori-ai-compactor - Multi-model reasoning consolidation
• midori-ai-context-bridge - Persistent thinking cache with decay
• midori-ai-mood-engine - Comprehensive mood management system
• midori-ai-reranker - Document reranking and filtering
• midori-ai-vector-manager - Vector storage abstraction

Media Management

• midori-ai-media-lifecycle - Time-based media decay and cleanup
• midori-ai-media-request - Type-safe media request/response interface
• midori-ai-media-vault - Encrypted media storage

Utilities & Examples

• midori-ai-agents-all - Meta-package with all packages and embedded docs
• midori-ai-agents-demo - Example implementations and demos